Estimating Missing Hydrological Data using Group Method of Data Handling (GMDH) in Mahanadi Basin, India

Abstract

Abstract High-quality hydrological data is essential for a wide range of applications, including the planning, design, operation, and maintenance of multipurpose water resource projects. It also plays a crucial role in utilizing various modelling and statistical methods for flood prediction and management, conducting hydrological analyses, estimating and monitoring environmental flows, as well as supporting research and development efforts. Hydrological data comprising of river Gauge (G), Discharge (D), Sediment (S) and Quality (Q) are collected at daily, weekly, ten-daily or monthly frequencies using either manual entry procedures or automatic measurement systems at hydrological observations sites. These are stored in databases that are made available to researchers, water managers etc for planning and research purposes. Missing data is a common problem in numerous hydrological databases that leads to inaccurate results, reducing statistical power and reliability of the data. Missing data also affects statistical analysis thereby reducing reliability and modelling conclusions drawn from using these incomplete datasets. To address this problem, the Group Method of Data Handling (GMDH) approach is applied to analyze data matrices from four hydrological observation sites in the Mahanadi River Basin, India. The daily discharge data is split into a learning set (70%) and a training set (30%) for the GMDH model. The Coefficients of determination (R2), Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) was used to evaluate the performance of GMDH model. The results illustrated the successful application of the GMDH algorithm in addressing missing data issues within discharge data, effectively filling the gaps.